Categorization

The categorization project's goal is to enable the agent to recognize objects and/or situations in the world and to divide the things he perceives in his environment into meaningful categories, such as plants, animals, cups, etc. There are many criteria for how objects can be categorized (a person for instance can be categorized as human, mammal, warm-blodded, two-legged, etc.), which makes categorization very difficult. Currently I am focusing only on the Structure of the objects the agent perceives.

A broad overview of work that has to be done

• Adaption of the hopfield net used to determine structure similarities to the microPSI environment, including several non-obvious design decisions, parameter tuning and implementation
• Creation of a fitting environment with which to work (very urgent!!)
• Testing of the general applicability of the idea using the above environment